I’m trying to run a simulation (A) with 3 radial subdomains using an initial condition obtained from another simulation (B), computed with a different radial resolution.
For that purpose, I tried to initialize A using a checkpoint from B (init_type = -1). When I do, I obtain the following behavior:
- If the 3 subdomains have an identical radial resolution in A and B, the simulation behaves as expected.
- If at least one of the subdomain of A has a resolution larger than the corresponding subdomain in B, all the output fields are filled with nan values, in the Python post-processing scripts.
- If at least one of the subdomain of A has a resolution smaller than the corresponding subdomain in B, all the output fields are filled with exact zeros, in the Python post-processing scripts.
Interestingly, if I use a single radial domain, the checkpoint initialization works properly if A has a larger resolution than B. However, the output fields remain filled with zeros if A has a smaller resolution than B.
Is it possible there is an issue with the interpolation routine, in particular if the simulation uses several subdomains. Do you have any idea on a potential fix or workaround?
Alternatively, I tried to initialize A using a generic initial condition produced from a 3D spherical output of B (init_type = 8). This method works as expected if I use a single radial domain. However, with several subdomains, the values are only mapped in the outer subdomain, independently of the resolution. The two remaining inner subdomains are filled with very small values (~1e-14).
Is the generic input file initialization compatible with a simulation using several subdomains?